MP15: Predictors of emergency department opioid use and variability of prescribing practices in a large multicenter Canadian cohort

Abstract

Introduction: Emergency department (ED) opioid prescribing has been linked to long-term use and dependence. Anecdotally, significant opioid practice variability exists between physicians and institutions, but this is poorly defined. Our objective was to collate and analyze multicenter data looking at predictors of ED opioid use and to identify potential areas for opioid stewardship. Methods: We linked administrative and computerized physician order entry (CPOE) data from all four ED's within our municipality over a one-year period. Eligible patients included those with a Canadian Triage and Acuity Scale (CTAS) pain complaint or an arrival numeric rating scale (NRS) pain score of greater than 3/10. Patients with missing demographic or chief complaint data were excluded. Multiple imputation was used for missing NRS pain scores. We performed descriptive analyses of opioid-treated and non-treated patients, followed by a multivariable logistic regression to identify predictors of ED opioid administration. Results: A total of 129,547 patients were studied. The mean age was 47.4 years and 55.4% were female. The median pain score was 6.6 in the no-opioid group and 8 in the opioid group. The most common pain categories were abdominal pain (23%), trauma (18.2%) and chest pain (15.3%). Overall, opioids were prescribed to 34% of patients. The most common CTAS score was CTAS 3 (44%), CTAS 1-2 42%) and CTAS 4-5 (13.9%). Multivariable predictors of opioid-use included the need for admission (adjusted OR 6.57; CI = 6.34-6.79), NRS pain score (aOR 1.24 per unit increase, CI 1.23-1.25), higher numerical CTAS score (aOR 0.89 per unit increase, CI 0.87-0.91), and chief complaints of back (aOR 7.69, CI 7.1-8.1), abdominal (aOR 5.9, CI 5.6-6.2), and flank pain (OR 3.8, CI 3.5-4). Oral opioids were prescribed in 39.8% of back pain presentations and 18.5% received IV opioids. Increasing age was a predictor but sex was not. There were significant institutional differences in opioid prescribing rates, with Hospital B being the least likely to prescribe opioids (aOR 0.82, CI 0.80-0.85) followed by Hospital C (aOR 0.83, CI 0.79-0.86) compared to the reference standard of Hospital A. Hospital D was most likely to prescribe opioids (aOR 1.32, CI 1.27-1.37). Conclusion: Predictors of ED opioid use were characterized using multicenter administrative data. Future research should seek to describe the physician- and site-level factors driving regional variation in opioid-based pain treatment.